Oral History of Gary Hendrix
Total Page:16
File Type:pdf, Size:1020Kb
Oral History of Gary Hendrix Interviewed by: Dag Spicer Recorded: November 19, 2004 Mountain View, California CHM Reference number: X3008.2005 © 2004 Computer History Museum Table of Contents NATURAL LANGUAGE UNDERSTANDING .........................................................................................................3 MACHINE TRANSLATION OF LANGUAGE..........................................................................................................7 ARTIFICIAL INTELLIGENCE....................................................................................................................................9 NATURAL LANGUAGE PROCESSING................................................................................................................11 SEMANTIC GRAMMARS........................................................................................................................................12 GENERALIZING SEMANTIC GRAMMARS .........................................................................................................14 ADVANCES IN ARTIFICIAL INTELLIGENCE......................................................................................................16 MICROCOMPUTERS ..............................................................................................................................................17 STARTING AN AI-BASED BUSINESS .................................................................................................................18 THE FIRST PROJECT.............................................................................................................................................19 MACHINE INTELLIGENCE CORPORATION ......................................................................................................21 STARTING SYMANTEC..........................................................................................................................................23 FINANCING THE COMPANY.................................................................................................................................25 DELIVERING A PRODUCT ....................................................................................................................................27 REORGANIZING SYMANTEC ...............................................................................................................................27 RESTARTING SYMANTEC ....................................................................................................................................29 REINVENTING SYMANTEC...................................................................................................................................31 OTHER SYMANTEC PRODUCTS ........................................................................................................................34 ACQUISITIONS ........................................................................................................................................................35 GOING PUBLIC ........................................................................................................................................................36 Gary Hendrix Conducted by Software History Center—Oral History Project Abstract: Gary Hendricks discusses in some detail his interest in natural language analysis and machine translations. He covers a number of his Artificial Intelligence related projects at SRI International. He describes how he and 15 other SRI employees formed the Machine Intelligence Corporation and the reasons why it didn’t succeed. He then goes into starting Symantec, how it was financed and the problems of having too many Ph.D.’s and the difficulties of trying to bring a more production-oriented approach to its offerings. He discusses Symantec’s initial products, its organization and its transformation into a publicly-held systems utilities company. This oral history ends with Gary Hendrix leaving Symantec in 1991. Dag Spicer: It’s Friday, November 19, 2004. I’m Dag Spicer and I’m here with Gary Hendrix at Mountain View, California at the Computer History Museum. Gary was the founder of Symantec in 1982. I just want to talk about two tracks with you. One is your professional disciplinary track which is computational linguistics and how you got interested in computing and the problems of language and language recognition, and then your experience with Symantec which you founded in 1982 and any lessons you might have learned from there. We’ll talk about numerous things. Does that sound all right? Gary Hendrix: It sounds like fun. Natural Language Understanding Spicer: Great. I’ll just throw out a date, 1970. ARPA came out with a series of funding for six different projects in natural language understanding, I believe. And this is about the time I think that you got into the field. Is that right? Hendrix: Well, that’s about right. I was an undergraduate at the University of Texas and I finished that up in 1970 in May and then I quickly got a Masters in computer science. That ended in December of 1970. And I was beginning to get into artificial intelligence just a little bit and sort of overlapping with operations research, but then beginning in January of 1971, I was CHM Ref: X3008.2005 © 2006 Computer History Museum Page 3 of 37 fulltime thinking about artificial intelligence and my supervising professor was Robert Simmons who was into computational linguistics and he was the man on the UT campus who was wired into that world. I also took a number of courses in other AI-related areas like robot planning, Laurent Siklossy, if I get his name right (he was always on my case about not saying it right), was another influence because he got me interested in robotics and, to some extent, the representation of semantic knowledge which I then used with Simmons and that’s where the name Symantec eventually came from. I did some work in the area of robotics and planning and I wrote a paper that I submitted for publication in one of the journals. And the people at SRI International read this paper and invited me out one summer. And there was a conference at Stanford. I guess it was one of the first joint conferences on artificial intelligence. And so because they had read this paper, they were interested in me and I eventually got hired by them even before I finished my Ph.D. Spicer: For people who don’t know what natural language understanding is, can you just tell us a little bit about what it is? It’s not speech recognition, is it? Hendrix: No, no. It’s trying to get computers to understand naturally-occurring languages like English or French or German as opposed to artificial languages like COBOL or C or FORTRAN or SQL, something like that. And there was a time when computers didn’t have graphical user interfaces and you had to communicate with them through some kind of language. You could argue that a graphical user interface is a kind of language too, but I’m talking about a more traditional language in which there are sentences of one kind or another. And we had thought that perhaps the use of English would open up computers for a wide range of uses and also, just as a science, it was and still is a fascinating field because we really don’t understand much about how the human mind works. We don’t really understand very much about how people process language and how they extract the meaning from it, how they think about things other than numerical kinds of things. And how you represent these in computers even thirty years later is still a pretty big mystery. Spicer: I’m very interested in the disciplinary underpinnings for that, which let’s just call it a disciplinary push versus say a contract pull which we’ll get to later. The U.S. Army had a great interest in voice-activated systems and so on for which your technology and interest would naturally meld. But in the theoretical foundations of computational linguistics, does that develop around Noam Chomsky and the early attempts in the late 1950s with artificial intelligence and the belief that we know a little bit about linguistics, enough to know that maybe we can extend it with a computer and produce really interesting research results, if not practical products? Hendrix: Well, certainly Chomsky was the big intellectual contributor to the field and he’s an absolute giant. There’s no one like Chomsky. And he developed many theories involving language. A lot of them were fairly mathematical in nature, describing how you could represent various kinds of grammar, how different levels of grammar behave mathematically. And people picked up on this and tried doing things with it. In one sense, the compilers are based on some CHM Ref: X3008.2005 © 2006 Computer History Museum Page 4 of 37 of Chomsky’s work in that the way you look at languages that you could use to design languages for machines. And what the people trying to understand natural languages were doing, well, they were saying, “Well, okay, we’re going to try to write a parser or we’re going to try to understand the grammar of English.” Computer languages are designed in such a way that once the syntax is understood, the meaning of what’s being expressed is also readily understood because the whole thing is artificial. It’s built up out of little pieces in the way computers are built up so the fundamental parts are all very well understood. I mean, if you get down to the bottom level, it’s ones and zeros, offs and ons. And then the components that are built up out of that are well understood